Database Reference
In-Depth Information
patient body mass indices may be abstracted by increasing-decreasing-
steady trend qualifiers. The authors also propose techniques for mining
such EHR temporal abstraction using standard data mining schemes
(e.g., apriori algorithm).
Neuvirth and his colleagues [35] proposed an interesting application of
data mining techniques on EHR data for the management of chronic dis-
eases. This application is able to predict patient future health states and
identify high risk patients for specific diseases where risk is a function of
the likelihood of needing emergency care and the likelihood of receiving
sub-optimal treatments. They further explore the links between physi-
cians treating these patient populations and outcomes to design a system
that optimizes the matching between individual patients and physicians
for better outcomes. Their analysis makes heavy use of standard ma-
chine learning techniques (e.g., logistic regression, K-Nearest Neighbor
classification) and survival analysis (Cox modeling) and has generated
interesting results for the management of diabetic patients.
The concept of patient similarity described above in Section 3.1.1 has
also been on EHR data with the AALIM system [36] which uses content
based search techniques on different modality data to extract disease
specific patient information and find groups of similar patients. AALIM
uses data from similar patients to help physicians make prognosis for
a given patient and design care management strategies. Sensor data
inputs into AALIM includes ECGs, videos, echocardiograms, MRIs and
text notes.
With the emergence of question answering systems like IBM Watson
[37], the potential to design systems able to ingest very large amounts of
structured and unstructured clinical data to support clinical diagnosis
and prognosis is emerging. The ability of Watson to analyze the meaning
and context of human language, and quickly process vast amounts of
information to answer questions has wide applicability in healthcare.
One can imagine applications where a properly trained Watson system
can assist decision makers, such as physicians and nurses, in identifying
the most likely diagnosis and treatment options for their patients. IBM
andWellpointhavepartneredtodevelopsuchasystemwithapplications
to patient diagnosis [38]. A similarly partnership with Memorial Sloan
Kettering is in place for the diagnosis and management of cancer [39]
4. Non-Clinical Healthcare Applications
The world is experiencing a rapid increase in its aging population,
and a corresponding increase in the prevalence of chronic diseases and
health care expenditure. For instance, the total Medicare expenditure in
